Professional Documents
Culture Documents
BOGNAREK - A Survey On Data Management Techniques in Military Wireless Sensor Networks - Hadmernök
BOGNAREK - A Survey On Data Management Techniques in Military Wireless Sensor Networks - Hadmernök
június
Abstract Absztrakt
INTRODUCTION
The term wireless sensor network refers to hundreds or even thousands of linked sensors that
can record changes in their physical environment. Acting in a coordinated fashion, these sensor
networks are able to continuously monitor the conditions of a particular area or physical object.
Recent technological developments make it possible to manufacture especially small and
inexpensive sensors that can be installed virtually anywhere. Due to their low cost and practical
versatility, these sensors have already become widespread in the civil sphere, with applications
in healthcare, manufacturing, environmental monitoring, and so-called smart environments.
They also have proven applications in military and defense, with sensors being placed on
soldiers’ bodies and in battlefield monitoring systems. The true potential of sensor networks in
all areas of life is only beginning to be explored. The further development of the Internet of
Things relies on research into sensor network technology.
However, there are major restrictions facing the optimal operation of sensor networks. These
restrictions stem from limitations in sensor memory, storage, computing capacity, and useful
life, as well as limitations on bandwidth. Therefore, efforts are being made to optimize the flow
of data traffic, routing between large numbers of nodes, and the storage and retrieval of data.
Any attempts at optimization must be carefully tailored according to scarcity of available
resources. The conventional data management solutions used in conventional computer
networks are somewhat inefficient, and because of this it is necessary to develop novel methods
and procedures. Several studies relevant to sensor network optimization have been published
domestically. The work of Nagy T. I. [1][2] focuses on routing strategies utilized in the network
layer of sensor networks. Haig Zs. [3][4] and Kovács L.’s [5] research explores potential
opportunities for use of sensor networks in electronic warfare. The ultimate goal of sensor
networks is the collection and storage of data, and to make that data available on demand.
Therefore, it is critical to understand the various processes through which these functions are
achieved.
Currently, researchers are furthering data management techniques that are better optimized
for large amounts of data resulting from sensor networks. Because the quantities of raw data
are so large, useful data must be specified and sorted according to their intended application.
To accomplish this aim, various data storage and query mechanisms are created. When
developing these new mechanisms, the limited storage capacity of the nodes must be taken into
consideration. Furthermore, there may be network communications and package delivery fees
associated with these mechanisms, which become financially significant. The purpose of this
article to suggest solutions for current data storage and query management issues by comparing
the advantages and disadvantages of different methods.
Since sensors have numerous limits in terms of storage capacity, communication capability,
and useful life, the development of novel procedures become necessary. One such procedure
leading to further optimization is the distributed processing and storage of data within the
network. By taking into consideration a number of specific parameters, we can optimize the
operation of the system and select the most effective method for data management. The
expected useful life of the system (including short and long-term installation), the nodes’ power,
computing and storage capacity, and the data requirements of the particular application are all
significant elements which must be considered. Furthermore, the nature of the user’s queries
(single, continuous, event-driven, etc.), the quantity of generated data during the operation, the
network size, and potential preprocessing parallelization opportunities are also all relevant.
Current research points towards the use of distributed data processing. There is a spatial and
temporal correlation between data collected by the sensors, which has substantial potential to
overcome many current limitations. The purpose of this chapter is to describe the characteristics
of sensor networks and the limitations of data management.
̶ during data transmission, longer delays may occur. It is not enough to optimize a
query once, but it must be continuously monitored and revised in order to operate at
the most efficient speed and resource utilization;
̶ since storage capacity of nodes is limited, the possibility of storing historical data
must be considered.
As an actively researched area, a number of solutions have been proposed which can optimize
the transmission routes between the nodes and reduce communication costs in the network.
During the processing of queries, a significant role is considered for the optimum operation of
the network layer and efficient routing algorithms. One such algorithm is known as directed
diffusion. To summarize, the initial request for data from the base station goes to all nodes,
flooding the network. Afterwards, the nodes generate gradients in order to identify where the
request is coming from and which neighboring nodes have the same data. When this type of
request is repeated, the network has already formed the most optimal path for processing it. If
queries are periodic, efficiency can be greatly increased. However, in the case of a new query,
it is necessary to start over and flood the network again. [7]
Data-centric storage
Between centralized and local storage, there is a middle course specifying that data must be
stored within network on predefined nodes according to the type of measurement and the place
of its origin. The central premise of the theory is to store the same data type on the same network
location. After that, any query that relates to a particular type of data is directly addressable to
that data’s storage, thereby avoiding queries associated with high communication costs. The
request can therefore be processed more efficiently, decreasing the energy and latency of
processing. Data storage in this system consists of two steps: If a sensor detects an event, it will
label it using a hash function; following this, the data will be directed to a node according to
the label. This process, GHT (geographic hash table), is a hash function that provides solutions
for mapping between the name and storage location of the event. Data centric-storage creates a
distributed storage structure, where events are named and grouped spatially. Names can be
regarded as hash function arbitrary keys and categorize basic units. This method is also
advantageous since the same data is stored in the same location. Through this, efficient
aggregation can be achieved. Of course, there is a critical question regarding this storage
strategy: How do we select the containing nodes? Additionally, for this strategy to function
effectively, the nodes must have increased storage capacity and resources. [13]
In order to make queries more effective, a hierarchy between storage nodes is formed.
The stores nearer to the base station have greater spatial and temporal coverage but are more
cumulative and contain compressed data. In contrast, storage in the lower levels is local and
contains more detailed data. Queries start from the base station and then reach the accumulated
data, only proceeding to lower levels of the hierarchy if greater accuracy is required. [10]
Because of the finite capacity of the storage nodes, algorithms are needed that can manage data
with the ability to delete or compress older values. Figure 2 describes the three storage
strategies:
network while using the least amount of resources possible, thereby increasing the useful life
of the network. This chapter will introduce different query interfaces which allow for access to
sensor network data, and techniques that optimize their efficiency.
system for query processing. Tsiftes N. and Dunkels described such a solution called Antelope
[19]. However, the more common option is for query interpretation, optimization, and transfer
to the data sources to remain on the client-side: users send SQL queries to the server, which
describe what kind of data the user interested in. Furthermore, users can specify how they want
data to be combined, transformed, summarized, and ultimately how they would like to receive
the query results. Using optimized routing strategies, the request is then transmitted to the
sensors. At the sensor level, some data preprocessing may happen: filtering undesirable values
and aggregating measurements. After this preprocessing, only the relevant information will be
transmitted to the base station, significantly reducing communication cost. The most widely
used implementation of the sensor database was developed on the TinyOS operating system
and is called TinyDB (Berkeley University). In the next chapter, the main features of sensor
database systems will be introduced according to the functions of TinyDB. [9][10][14]
TinyDB sensor database architecture and data model
In terms of architecture, it is necessary to separate operations on the server side from that of the
sensors during query processing. In close collaboration, the two sides work together to ensure
the efficient execution of queries. With the help of base station applications, we can formulate
queries and understand SQL code interpretation. The goal is to optimize transmission of data
towards the sensors using the most efficient routing protocol. For the sensors, there is a query
processing subsystem running for each device on the TinyOS operating system. It receives
queries on the nodes, optimizes their runtime, and maintains a catalog consisting of different
metadata. The system offers the opportunity for data aggregation, energy efficient transmission,
and cooperation between the different layers, which will be introduced in the following section
based on the works of Madden, S. et al. [14][15] The structure of this system is shown in Figure
4.
The database-oriented approach has two basic objectives: making wording similar to that of
conventional database systems thereby allowing simple SQL query commands, and to minimize
the energy consumption of the network during the data collection process. [8]
In order to make simple SQL queries, it is useful to imagine a logic level virtual relational table
above the physical distributed system in which the columns contain attributes (time stamp, node
ID, values etc.) and the rows represent individual measurements.
Sensors table
… … … … …
As mentioned previously, in addition to the traditional query types, sensor networks must
also handle queries on data streams. The already known SELECT column1, column2, sensors
FROM table [WHERE condition] [GROUP BY column1 [column2]] [HAVING group
condition] command is therefore extended to the [PERIOD SAMPLE FOR duration] or
[LIFETIME period] clause, which specifies the data collection frequency and duration,
allowing for continuous queries of data streams. Also, the [STOP ON EVENT (parameter) and
WHERE (parameter)] clauses can be used to allow for event-based suspension of queries, and
the [OUTPUT ACTION (event)] clause can be used to specify an event that is executed if the
statement conditions are true. Of course, various aggregation functions are also supported:
(AVG, MIN, MAX, COUNT, SUM).
calculation of the overall aggregation. Some nodes summarize the values obtained from nodes
in lower levels, so that the base station only receives aggregated values.
Figure 5 TinyDB aggregation within the network cooperation between layers [15]
database technologies (e.g. MongoDB1, Apache Cassandra2, HBase3 etc.). Considering the data
needs of sensor networks, it is more optimal to use NoSQL databases in several respects.
Collection is done by modifying data needs, as opposed to more complex queries that are part
of relational database management. Therefore, NoSQL databases also most ideal regarding
speed of query processing, in particular for continuously generated data streams in large
amounts.
SUMMARY
In selecting a suitable data management strategy, the issues of storage and query processing go
hand in hand. Deep knowledge of the different storage models and related query processing
solutions is needed. Addressing data management strategy is essential for further developments
in the field. Table 2. summarizes the strategies and relative merits of the various data storage
systems.
Communication
between nodes and High Low High
the base station
Communication
between nodes Low High Low
processing queries
Optimal quantity of
Small Medium Large
data being treated
Homogeneous and
Optimal data type Mostly homogeneous Mostly homogeneous
heterogeneous
Compromising gateway,
Compromising gateway, Less
Security Challenges Compromising gateway Node vulnerability, Data
control over data
manipulation
Sensor distributed
Relational and databases, SQL-based Distributed databases in the
Query Management
traditional databases query processing (e.g. cloud
TinyDB)
Continuously running
Event-driven, Ad hoc queries for extended Past data for queries;
Optimal query types "Snapshot," Continuously
queries on past data times, Multidimensional
queries for domain running, Event-driven and/or
1
https://www.mongodb.com/
2
http://cassandra.apache.org/
3
https://hbase.apache.org/
Processing resource
Server Server and nodes Theoretically unlimited
constraints
Table 2. Comparison of sensor networks and storage strategies (by the author)
CONCLUSIONS
Due to the low cost and practical versatility sensors have already become widespread in the
civil sphere and they also have proven applications in military and defense. The major sensor
network data management solutions demonstrate that the range of possibilities is quite broad.
Presently, traditional centralized storage is standard; however, as sensor network data becomes
more complex and sensor size shrinks, new solutions are required. Ongoing research in this
field points towards distributed data management within the network. A major goal of this
research is to optimize the energy consumption of nodes. This is best achieved by using efficient
routing strategies that enable communication between nodes and greater collaboration between
protocol layers. As the role of Big Data and the Internet of Things grows, local resources will
become obsolete and eventually replaced by more solutions with a greater degree of specificity
and customization. Improvements in hosting and cloud technologies will render these new
network solutions nearly unlimited in their scope. Advanced data analysis solutions can provide
computing capacity on demand. In the process of selecting a particular technology, it is
important to consider the needs of the application or user. It is relevant to consider the type of
data being stored, retrieved, analyzed and selected. Furthermore, it is essential to choose the
most appropriate and efficient storage and query-processing procedures.
BIBLIOGRAPHY
[1] NAGY T. I.: Hálózati réteg a szenzorhálózatokban; Hadmérnök, VII 3 (2012) pp. 123–
130.
[2] NAGY T. I.: A felügyelet nélküli szenzorhálózatok és a programozási nyelvek
kapcsolata; Hadmérnök, IV 4 (2009) pp. 303–311.
[3] HAIG ZS.: Networked unattended ground sensors for battlefield visualization;
AARMS, 3 3 (2004) pp. 387–399.
[4] HAIG ZS.: Network-Centric Warfare and sensor fusion; AARMS, 2 2 (2003) pp. 245–
256.
[5] KOVÁCS L.: Az elektronikai felderítés korszerű eszközei, eljárásai és azok
alkalmazhatósága a Magyar Honvédségben. Budapest: ZMNE, 2003. (PhD értekezés)
[6] ZHAO, F., GUIBAS, J. L.: Wireless Sensor Networks: An Information Processing
Approach; Morgan Kaufmann 2004.
[7] FAN, W., HAO, Z., GUO, Z.: Wireless Sensor Network Data Storage Optimization
Strategy In: WANG, X. et al.: Advanced Technologies in Ad Hoc and Sensor Networks,
Proceedings of the 7th China Conference on Wireless Sensor Networks; Springer 2014.
pp. 345-351.
[8] RAWAT, P. et al.: Wireless sensor networks: a survey on recent developments and
potential synergies; The Journal of Supercomputing 68. 1. (2014) pp. 1–48.
[9] WEI-PENG, C., JENNIFER C. H.: Data Gathering and Fusion in Sensor Networks In:
IVAN, S.: Handbook of Sensor Networks: Algorithms and Architectures; John Wiley &
Sons 2005. pp. 493-526.
[10] TRIGONI, N. et al.: Querying of Sensor Data In: GAMA, J., GABER, M. M. (Eds.):
Learning from Data Streams, Processing Techniques in Sensor Networks; Springer
2007. pp. 73-84.
[11] DIALLO, O. et al.: Distributed Database Management Techniques for Wireless Sensor
Networks; IEEE Transactions on Parallel and Distributed Systems 26. 2. (2015) pp.
604-620.
[12] JALLAD, A-H., VLADIMIROVA, T.: Data-Centricity in Wireless Sensor Networks In:
MISRA, C., S., WOUNGANG, I., MISRA, S.: Guide to Wireless Sensor Networks.
Computer Communications and Networks; Springer 2009. pp. 183-204.
[13] GOVINDAN, R.: Data-Centric Routing and Storage in Sensor Networks In:
RAGHAVENDRA, C. S. et al.: Wireless Sensor Networks; Springer 2004. pp. 185-205.
[14] GEHRKE, J., MADDEN, S.: Query processing in sensor networks; IEEE Pervasive
Computing, 3. 1. (2004) pp. 46-55.
[15] MADDEN, S. et al: TinyDB: an acquisitional query processing system for sensor
networks; ACM Transactions on Database Systems, 30. 1. (2005) pp. 122-173.
[16] VAN DER VEEN, J. S. et al.: Sensor Data Storage Performance: SQL or NoSQL,
Physical or Virtual; 2012 IEEE Fifth International Conference on Cloud Computing,
2012, pp. 431-438.
[17] MISTRAL SOLUTIONS: The Sensor Cloud and Homeland Security; Online:
https://www.mistralsolutions.com/sensor-cloud-homeland-security/ (Letöltve: 2017. 11.
26.)
[18] CAO, H. et al.: Cloud-Assisted UAV Data Collection for Multiple Emerging Events in
Distributed WSNs; Sensors, 17. 8. (2017), Online: http://www.mdpi.com/1424-
8220/17/8/1818 (Letöltve: 2017. 11. 26.)
[19] TSIFTES, N., DUNKELS, A.: A database in every sensor; SenSys '11, Proceedings of
the 9th ACM Conference on Embedded Networked Sensor Systems, 2011, pp. 316-332.
[20] LANGSON, J.: Security Implications of Data Dissemination Methods in Wireless
Sensor Networks, Online:
https://pdfs.semanticscholar.org/9734/15286a035d2a541d7599e7faf06512229324.pdf
(Letöltve: 2017. 11. 26.)
[21] WINKLER, M., TUCHS, K. D., HUGHES, K., BARCLAY, G.: Theoretical and
Practical aspects of military wireless sensor networks. Journal of Telecommunications
and Information Technology, 2 1 (2008), 37–45.